Optimized Framework for Modeling Agent Based Distributed System

Abstract

The advancement in social networks has bought a lot of changes in the data mining community. In the social networks, sentiment analysis supports as base for recommendation systems to extract the information on user emotional state to improve user satisfaction. Sentiment analysis aims at sorting and detecting the polarity of an input text, sentence, or document Due to the vast amount of data in the social network, obtaining required services and accurate recommendation was not performed with less time. Therefore, three novel methods are developed to enhance the recommendation system in social network by using multi agent. newlineIn the first research work, Hessian Distributed Ant Optimized and Perron Frobenius Eigen Centrality (HDAO-PFEC) multi agent based method is designed to analyze sentiments in social networks. The main aim of designing HDAO-PFEC method is to reduce the running time and data storage overhead in an accurate manner. Hessian Mutual Distributed Ant Optimization model is employed in map phase to gain similar user interest tweets. Then the Perron Frobenius Eigen Vector Centrality model is applied in reduce phase to obtain accurate and dimensionality reduced tweets. The proposed HDAO-PFEC method effectively improves the accuracy score by 13% and 20% as compared existing Laplace three level stochastic variational inference and multi-agent based distributed architecture respectively. Also, running time of proposed HDAO-PFEC method is reduced by 21% and 29% as compared existing methods respectively. In addition, data storage overhead is decreased by 27% and 42% when compared existing methods. newlineIn the second research work, Friedman and Frobenius Matrix Collaborative Recommendation (F-FMCR) method is proposed to generate efficient recommendation system in social network. Proposed F-FMCR method is introduced by combining MapReduce Friedman function and Matrix Collaborative (FMC) function. Most relevant tweets are initially selected in F-FMCR with the help of MapReduce Friedman function in the multi-agent system.